[off-topic] Re: [xml-dev] Use xsd to specify multiple instances ofexisti

[Rick Jelliffe <rjelliffe@allette.com.au>]

Michael Kay wrote:
> Or you could do it all with assertions - but the danger with complex
> assertions is that the diagnostics when they are not satisfied can get
> rather obscure.
>   
Hence the criticality of natural language formulations of the intent of 
the assertion test. Which was the
departure point for Schematron in 1999.

The initial problem that computer scientists try to solve is how to 
represent something in a way that
computers can understand. Then, as soon as that is answered, they 
realize that they have asked the
wrong question, and they need to ask how to represent something in ways 
that both humans and
computers can understand. I do think XSD is stuck asking that first 
question, and never getting to
the second (or deferring it to implementations, as if good human text 
about the intentions and rationale
of some constraint can be auto-generated from machine descriptions.)

This relates to the patterns versus types point I tried to make the 
other day. Considering a document
as a graph of nodes and arcs, XSD allows you to document nodes but not 
arcs. So there is nowhere
to document why it is that two elements in a sequence should follow each 
other, for example. That
becomes a comment of the type or parent, but even there it is odd. I 
think Schematron is equally strong
(and equally weak) on documentation and constraints on arcs as well as 
nodes.

Cheers
Rick